Antiferromagnetic Kitaev Model

• Antiferromagnetic Kitaev model is a bond-directional spin system characterized by positive Kitaev couplings on a honeycomb lattice and an exact solution via Majorana partonization.
• It exhibits distinct magnetic behavior under applied fields, where chirality terms induce gap openings and intermediate phases that break the AFM–FM equivalence.
• The model reveals symmetry-enriched vison dynamics with projective translations, leading to nontrivial thermal Hall effects and potential for exotic doped phases.

The antiferromagnetic Kitaev model is the bond-directional spin model in which the nearest-neighbor Kitaev couplings are positive, conventionally written on the honeycomb lattice as

$$H_K=\sum_{\langle ij\rangle_\gamma} K_\gamma S_i^\gamma S_j^\gamma,\qquad \gamma\in\{x,y,z\},\qquad K_\gamma>0,$$

with the isotropic case $K_x=K_y=K_z\equiv K>0$ serving as the canonical reference point. In the pure nearest-neighbor honeycomb problem, the antiferromagnetic and ferromagnetic signs are formally related by a sublattice-dependent spin rotation, so the zero-field spectra coincide; however, this equivalence is lifted by magnetic fields, further-neighbor and off-diagonal exchanges, anisotropy, and doping, and the antiferromagnetic model then exhibits a distinct combination of field robustness, intermediate phases, projective vison symmetry, and topological transport signatures (Yu et al., 2012, Sugita et al., 2019, Chen et al., 2022).

1. Hamiltonian structure, exact solution, and zero-field equivalence

For the spin-$1/2$ honeycomb model, the exact solution proceeds through Majorana partonization,

$$S_i^\gamma=\frac{i}{2}b_i^\gamma c_i,\qquad u_{ij}=i b_i^\gamma b_j^\gamma,$$

with $u_{ij}=\pm1$ static $\mathbb{Z}_2$ link variables on a $\gamma$-bond and plaquette flux

$W_p=\prod_{\langle ij\rangle\in p} u_{ij}.$

A vison is the $\pi$-flux excitation with $W_p=-1$. In the exactly solvable limit the $K_x=K_y=K_z\equiv K>0$0 and $K_x=K_y=K_z\equiv K>0$1 commute with the Kitaev Hamiltonian, so visons are static and the itinerant Majorana fermions propagate in a static $K_x=K_y=K_z\equiv K>0$2 gauge background (Chen et al., 2022).

At $K_x=K_y=K_z\equiv K>0$3, the isotropic honeycomb model belongs to the gapless Kitaev spin-liquid regime; more generally, for fixed $K_x=K_y=K_z\equiv K>0$4, the zero-field phase diagram contains gapped Abelian $K_x=K_y=K_z\equiv K>0$5 phases when $K_x=K_y=K_z\equiv K>0$6 and otherwise the gapless Kitaev spin liquid (Holdhusen et al., 2023). The antiferromagnetic sign does not alter the pure-model spectrum on the bipartite honeycomb lattice because a sublattice-dependent $K_x=K_y=K_z\equiv K>0$7-rotation maps $K_x=K_y=K_z\equiv K>0$8. In the $K_x=K_y=K_z\equiv K>0$9–$1/2$0 construction, the pure Kitaev line occurs at $1/2$1, where $1/2$2, so $1/2$3 realizes the antiferromagnetic Kitaev coupling (Yu et al., 2012). The same zero-field equivalence is emphasized in microscopic derivations for polar Ru trihalides and $1/2$4-electron honeycombs, but those works also stress that realistic $1/2$5, $1/2$6, $1/2$7, DM, and longer-range exchanges immediately make the sign of $1/2$8 observable in the phase diagram and field response (Sugita et al., 2019, Jang et al., 2018).

The higher-spin generalizations retain key Kitaev diagnostics but lose exact solvability. Tensor-network calculations for $1/2$9 found $$S_i^\gamma=\frac{i}{2}b_i^\gamma c_i,\qquad u_{ij}=i b_i^\gamma b_j^\gamma,$$0 and vanishing spin-spin correlations beyond nearest neighbors in the low-field regime, while DMRG for the $$S_i^\gamma=\frac{i}{2}b_i^\gamma c_i,\qquad u_{ij}=i b_i^\gamma b_j^\gamma,$$1 honeycomb model identified a $$S_i^\gamma=\frac{i}{2}b_i^\gamma c_i,\qquad u_{ij}=i b_i^\gamma b_j^\gamma,$$2 gauge structure with uniform plaquette flux $$S_i^\gamma=\frac{i}{2}b_i^\gamma c_i,\qquad u_{ij}=i b_i^\gamma b_j^\gamma,$$3 and topological loop sectors on cylinders (Jahromi et al., 2021, Khait et al., 2020).

2. Magnetic fields, intermediate phases, and the breakdown of AFM–FM equivalence

A weak $$S_i^\gamma=\frac{i}{2}b_i^\gamma c_i,\qquad u_{ij}=i b_i^\gamma b_j^\gamma,$$4 field generates the chirality term

$$S_i^\gamma=\frac{i}{2}b_i^\gamma c_i,\qquad u_{ij}=i b_i^\gamma b_j^\gamma,$$5

which gaps the gapless Majorana spectrum into the non-Abelian Ising phase. DMRG and ED found that this low-field chiral spin liquid remains stable in the antiferromagnetic honeycomb model up to $$S_i^\gamma=\frac{i}{2}b_i^\gamma c_i,\qquad u_{ij}=i b_i^\gamma b_j^\gamma,$$6, whereas the corresponding ferromagnetic phase is lost already near $$S_i^\gamma=\frac{i}{2}b_i^\gamma c_i,\qquad u_{ij}=i b_i^\gamma b_j^\gamma,$$7; for $$S_i^\gamma=\frac{i}{2}b_i^\gamma c_i,\qquad u_{ij}=i b_i^\gamma b_j^\gamma,$$8, the same study reported an intermediate gapless phase before the high-field partially polarized paramagnet (Zhu et al., 2017). A Majorana mean-field study of $$S_i^\gamma=\frac{i}{2}b_i^\gamma c_i,\qquad u_{ij}=i b_i^\gamma b_j^\gamma,$$9 fields across several two- and three-dimensional Kitaev lattices generalized this asymmetry: antiferromagnetic couplings generically produced substantially larger critical fields and an intermediate spin liquid induced by a field-driven sign change in an effective $u_{ij}=\pm1$0-bond energy parameter, while the ferromagnetic models polarized directly (Yang et al., 2020).

The detailed structure of the intermediate $u_{ij}=\pm1$1-field regime remains method-sensitive. Hierarchical mean-field theory with ED benchmarking on the antiferromagnetic honeycomb model resolved four regimes as $u_{ij}=\pm1$2 increases: a low-field KSL with many-body Chern number $u_{ij}=\pm1$3, a stripe-ordered phase entered at $u_{ij}=\pm1$4, a chiral partially polarized phase entered at $u_{ij}=\pm1$5, and a trivial partially polarized phase above $u_{ij}=\pm1$6 (Holdhusen et al., 2023). By contrast, earlier DMRG/ED work interpreted the same field window as a single intermediate gapless phase (Zhu et al., 2017). A plausible implication is that the antiferromagnetic intermediate regime is unusually susceptible to the choice of variational manifold, cluster geometry, and finite-size scaling.

Field orientation adds a further layer of structure. For arbitrary orientations, Majorana mean-field theory for the pure antiferromagnetic honeycomb model found along $u_{ij}=\pm1$7 the Chern-number sequence $u_{ij}=\pm1$8 with band touchings near $u_{ij}=\pm1$9 and $\mathbb{Z}_2$0, while along $\mathbb{Z}_2$1 it found $\mathbb{Z}_2$2 in the corresponding field range. In the $\mathbb{Z}_2$3–$\mathbb{Z}_2$4 plane, the Chern number changes sign across the line $\mathbb{Z}_2$5, equivalently $\mathbb{Z}_2$6, reflecting the sign of $\mathbb{Z}_2$7 (Yılmaz et al., 2022).

The $\mathbb{Z}_2$8 problem also exhibits a sharp AFM–FM contrast in fully correlated variational calculations. Using a Jordan–Wigner fermionization with a generalized BCS plus Jastrow wavefunction, the antiferromagnetic honeycomb model showed two transitions, a continuous one at $\mathbb{Z}_2$9 and a first-order one at $\gamma$0, separated by a gapless intermediate state with fluctuating $\gamma$1 fluxes. The ferromagnetic case instead displayed a single first-order transition at $\gamma$2 (Ido et al., 2019).

3. Vison symmetry enrichment, projective translations, and vison Hall response

The antiferromagnetic honeycomb model develops a qualitatively different vison sector once a Zeeman field mobilizes fluxes. In the solvable reference problem used to regularize this dynamics,

$\gamma$3

the three-spin term gaps the itinerant Majoranas and yields exponentially localized visons. Local $\gamma$4 operators generated by the Zeeman term flip adjacent $\gamma$5 variables and thereby move a vison across a bond. The loop phases accumulated by these hopping amplitudes diagnose whether lattice translations act projectively on the vison (Chen et al., 2022).

The key distinction is that the antiferromagnetic vison sees $\gamma$6 flux per Bravais unit cell. Equivalently, its translation operators obey the magnetic-translation algebra

$\gamma$7

whereas the ferromagnetic vison has ordinary commuting translations. This difference is symmetry-enriched rather than merely energetic: the ferromagnetic and antiferromagnetic honeycomb models realize distinct symmetry-enriched Ising topological orders (Chen et al., 2022).

Property	AFM Kitaev model	FM Kitaev model
Vison unit-cell phase	$\gamma$8	$\gamma$9
Translation algebra	$W_p=\prod_{\langle ij\rangle\in p} u_{ij}.$0	$W_p=\prod_{\langle ij\rangle\in p} u_{ij}.$1
Vison bands	Two gapped bands with $W_p=\prod_{\langle ij\rangle\in p} u_{ij}.$2	One band with $W_p=\prod_{\langle ij\rangle\in p} u_{ij}.$3
Intrinsic vison $W_p=\prod_{\langle ij\rangle\in p} u_{ij}.$4	Nonzero in principle	Zero

In the antiferromagnetic case, projective translation doubles the vison unit cell and produces two gapped bands with opposite Chern numbers,

$W_p=\prod_{\langle ij\rangle\in p} u_{ij}.$5

whereas the ferromagnetic case yields a single band with identically vanishing Berry curvature. Representative scales reported for the antiferromagnetic regime include $W_p=\prod_{\langle ij\rangle\in p} u_{ij}.$6 at $W_p=\prod_{\langle ij\rangle\in p} u_{ij}.$7, and a weak-field hopping amplitude

$W_p=\prod_{\langle ij\rangle\in p} u_{ij}.$8

which is parametrically suppressed relative to the ferromagnetic scaling $W_p=\prod_{\langle ij\rangle\in p} u_{ij}.$9 (Chen et al., 2022).

This band topology feeds directly into transport. Treating visons as dilute hard-core bosons in topological bands, the intrinsic thermal Hall conductivity takes the standard bosonic-band form

$\pi$0

Because $\pi$1 in the ferromagnetic vison band, $\pi$2 there. In the antiferromagnetic model, $\pi$3 and the thermal population imbalance between the $\pi$4 bands generates a nonzero intrinsic vison contribution to $\pi$5 (Chen et al., 2022).

4. High-field polarized phase, dressed quasiparticles, and dynamical structure factor

In the large-$\pi$6-field regime, the antiferromagnetic honeycomb model is conveniently analyzed from the fully polarized state using perturbative continuous unitary transformations. After rotating the spin basis so that the field points along the new $\pi$7-axis, the elementary excitations are dressed spin flips represented as hard-core bosonic quasiparticles,

$\pi$8

The pCUT construction yields an effective Hamiltonian $\pi$9 that conserves quasiparticle number; in the cited calculation, the 1QP Hamiltonian was computed to order 8 and the 2QP Hamiltonian to order 7 (Schellenberger et al., 2022).

The high-field polarized phase persists up to a critical ratio $W_p=-1$0, with method-dependent estimates $W_p=-1$1 from pCUT, $W_p=-1$2 from tensor networks, and $W_p=-1$3 from DMRG. Within this phase the one-particle sector contains two magnon-like bands $W_p=-1$4 and $W_p=-1$5, both softened by increasing $W_p=-1$6, with maximal splitting at $W_p=-1$7. A parity selection rule suppresses the lower-band intensity at $W_p=-1$8 in the dynamical structure factor (Schellenberger et al., 2022).

The two-particle sector is not a featureless continuum. Three two-quasiparticle continua arise from $W_p=-1$9, but a comparatively strong spectral feature appears above the upper edge of the highest continuum: three antibound states generated primarily by repulsive nearest-neighbor density-density interactions. These states are dominated by nearest-neighbor relative separations, lie just above the highest continuum, and remain nearly flat in momentum space (Schellenberger et al., 2022).

Quasiparticle decay sets in when the upper 1QP band overlaps the lowest 2QP continuum. In the free-particle analysis this occurs for $K_x=K_y=K_z\equiv K>0$00, after which the upper one-particle mode acquires a finite lifetime and its spectral weight broadens. The lower 1QP band does not intersect the 2QP continua in the investigated regime and shows no corresponding decay signature (Schellenberger et al., 2022).

5. Higher spin, anisotropy, and doped descendants

The antiferromagnetic Kitaev phenomenology extends beyond the exactly solvable $K_x=K_y=K_z\equiv K>0$01 isotropic honeycomb point. For the $K_x=K_y=K_z\equiv K>0$02 honeycomb model, DMRG found strong evidence for a $K_x=K_y=K_z\equiv K>0$03 spin liquid at zero field, with strictly nearest-neighbor bond-directional correlations and an upper bound on the excitation gap of $K_x=K_y=K_z\equiv K>0$04 on the largest $K_x=K_y=K_z\equiv K>0$05 cylinders studied. Under a $K_x=K_y=K_z\equiv K>0$06 field, the $K_x=K_y=K_z\equiv K>0$07 geometry displayed an intermediate gapless quantum liquid between $K_x=K_y=K_z\equiv K>0$08 and $K_x=K_y=K_z\equiv K>0$09, whereas the ferromagnetic $K_x=K_y=K_z\equiv K>0$10 model polarized directly near $K_x=K_y=K_z\equiv K>0$11 (Khait et al., 2020).

For general spin $K_x=K_y=K_z\equiv K>0$12, tensor-network calculations combined with high-field linked-cluster expansions found a flux-free low-field state for all $K_x=K_y=K_z\equiv K>0$13, but also a discrete orientational symmetry breaking at $K_x=K_y=K_z\equiv K>0$14 for $K_x=K_y=K_z\equiv K>0$15, consistent with dimerized bond-energy patterns. The same study identified an intermediate region for every spin value considered, with the number of distinct subregions increasing with $K_x=K_y=K_z\equiv K>0$16, and argued that the collapse of the high-field polarized phase is unconventional because the one-particle gap closes very flatly while the gap-mode spectral weight remains finite for $K_x=K_y=K_z\equiv K>0$17. In the classical limit the critical line approaches $K_x=K_y=K_z\equiv K>0$18 (Jahromi et al., 2021).

Strong bond anisotropy exposes a different facet of antiferromagnetic Kitaev order. In the toric-code limit $K_x=K_y=K_z\equiv K>0$19, the antiferromagnetic Kitaev–Heisenberg–$K_x=K_y=K_z\equiv K>0$20 magnet maps onto a toric-code Hamiltonian with

$K_x=K_y=K_z\equiv K>0$21

a topological entanglement entropy $K_x=K_y=K_z\equiv K>0$22 in the $K_x=K_y=K_z\equiv K>0$23 spin liquid, and symmetry-enriched low-energy $K_x=K_y=K_z\equiv K>0$24-doublets. In this limit, the QSL-to-Heisenberg-ordered transition is continuous and described by a self-dual modified Abelian Higgs theory with mutual Chern–Simons coupling, whereas the QSL-to-$K_x=K_y=K_z\equiv K>0$25 paramagnet transition is first order (Nanda et al., 2021).

Carrier doping removes the AFM–FM equivalence even more decisively. In an SU(2) slave-boson mean-field treatment of the doped honeycomb Kitaev–Heisenberg model, the antiferromagnetic Kitaev side first supports a time-reversal-breaking chiral triplet $K_x=K_y=K_z\equiv K>0$26-wave superconductor $K_x=K_y=K_z\equiv K>0$27SC$K_x=K_y=K_z\equiv K>0$28, and then, with further doping, a chiral singlet $K_x=K_y=K_z\equiv K>0$29 state. When antiferromagnetic Kitaev and ferromagnetic Heisenberg interactions compete, a distinct time-reversal-symmetric triplet $K_x=K_y=K_z\equiv K>0$30-wave state $K_x=K_y=K_z\equiv K>0$31SC$K_x=K_y=K_z\equiv K>0$32 appears; in that phase a bulk gap closing near $K_x=K_y=K_z\equiv K>0$33 marks a transition from an odd-parity trivial state to an odd-parity topological state in class DIII (Okamoto, 2012).

6. Microscopic routes and candidate materials

In conventional low-spin $K_x=K_y=K_z\equiv K>0$34 edge-sharing octahedra, the Jackeli–Khaliullin mechanism typically produces a ferromagnetic Kitaev exchange $K_x=K_y=K_z\equiv K>0$35. A central development in the antiferromagnetic literature is the identification of microscopic situations where this sign is reversed. In polar Ru hydrides $K_x=K_y=K_z\equiv K>0$36-RuH$K_x=K_y=K_z\equiv K>0$37 ($K_x=K_y=K_z\equiv K>0$38Cl, Br), broken inversion symmetry unbalances the two ligand-mediated paths and activates a $K_x=K_y=K_z\equiv K>0$39 superexchange channel giving

$K_x=K_y=K_z\equiv K>0$40

The ab initio exchanges extracted for these monolayers were $K_x=K_y=K_z\equiv K>0$41 meV, $K_x=K_y=K_z\equiv K>0$42 meV for $K_x=K_y=K_z\equiv K>0$43-RuH$K_x=K_y=K_z\equiv K>0$44Cl$K_x=K_y=K_z\equiv K>0$45, and $K_x=K_y=K_z\equiv K>0$46 meV, $K_x=K_y=K_z\equiv K>0$47 meV for $K_x=K_y=K_z\equiv K>0$48-RuH$K_x=K_y=K_z\equiv K>0$49Br$K_x=K_y=K_z\equiv K>0$50, i.e. dominant antiferromagnetic Kitaev couplings substantially larger than the ferromagnetic $K_x=K_y=K_z\equiv K>0$51 meV quoted for $K_x=K_y=K_z\equiv K>0$52-RuCl$K_x=K_y=K_z\equiv K>0$53 in the same analysis (Sugita et al., 2019).

A distinct AFM route arises in $K_x=K_y=K_z\equiv K>0$54-electron honeycombs. For $K_x=K_y=K_z\equiv K>0$55PrO$K_x=K_y=K_z\equiv K>0$56 ($K_x=K_y=K_z\equiv K>0$57Li, Na), Pr$K_x=K_y=K_z\equiv K>0$58 realizes a $K_x=K_y=K_z\equiv K>0$59 Kramers doublet behaving as $K_x=K_y=K_z\equiv K>0$60, and oxygen-mediated superexchange through the $K_x=K_y=K_z\equiv K>0$61–$K_x=K_y=K_z\equiv K>0$62–$K_x=K_y=K_z\equiv K>0$63 and $K_x=K_y=K_z\equiv K>0$64–$K_x=K_y=K_z\equiv K>0$65–$K_x=K_y=K_z\equiv K>0$66 channels produces a dominant antiferromagnetic Kitaev interaction with smaller $K_x=K_y=K_z\equiv K>0$67, $K_x=K_y=K_z\equiv K>0$68, and $K_x=K_y=K_z\equiv K>0$69. The calculations identified Li$K_x=K_y=K_z\equiv K>0$70PrO$K_x=K_y=K_z\equiv K>0$71 as the most promising member of the series (Jang et al., 2018).

Experimental proximity to antiferromagnetic Kitaev physics has also been discussed for high-spin $K_x=K_y=K_z\equiv K>0$72 cobaltates. In Li$K_x=K_y=K_z\equiv K>0$73Co$K_x=K_y=K_z\equiv K>0$74SbO$K_x=K_y=K_z\equiv K>0$75, neutron diffraction and thermodynamics found zero-field A-type antiferromagnetic order below $K_x=K_y=K_z\equiv K>0$76 K, a recovered magnetic entropy of $K_x=K_y=K_z\equiv K>0$77 between 2 K and 50 K, and a spin-flop-driven crossover to ferromagnetic order near $K_x=K_y=K_z\equiv K>0$78 T at 2 K. The interpretation given was that antiferromagnetic Kitaev exchange coexists with ferromagnetic Heisenberg exchange and off-diagonal anisotropies, placing the material near a Kitaev quantum-spin-liquid sector rather than within a pure Kitaev phase (Vivanco et al., 2020).

The antiferromagnetic sign alone does not guarantee a spin liquid. On the triangular lattice, first-principles work on NaRuO$K_x=K_y=K_z\equiv K>0$79 derived a $K_x=K_y=K_z\equiv K>0$80 model with a notably antiferromagnetic Kitaev term but dominant positive $K_x=K_y=K_z\equiv K>0$81 and ferromagnetic $K_x=K_y=K_z\equiv K>0$82, plus strongly anisotropic four-spin ring exchange. Classical minimization and exact diagonalization both pointed instead to a robust easy-plane ferromagnetic order (Razpopov et al., 2022). A plausible implication is that AFM Kitaev exchange enlarges the accessible parameter space for frustrated quantum magnetism, but the realized phase remains controlled by the full tensorial exchange structure rather than by the sign of $K_x=K_y=K_z\equiv K>0$83 in isolation.